Glancing at Extended Reality: An Empirical Model of 3D Animated XR Data Traffic
Tiziana Cattai, Luca Mastrandrea, Alessandro Priviero, Gaetano Scarano, Stefania Colonnese · 2024
Several elements in the design of next generation networks, such as user profiling and network slicing, are based on precise models of the traffic load. In this context, recent research has investigated various video traffic classes, while traffic related to extended reality (XR) services remains unexplored. In this paper, we propose an original empirical model of 3D animated XR data, derived by encoding real point clouds with a standard compliant codec. Our proposed approach spans different temporal scales, from minutes to milliseconds aiming to measure different phenomena. Indeed, we firstly analyze the packet size distribution at the lower time scale, and we identify that it is well approximated by heavy tailed Gamma distribution. Then, we demonstrate how this finding can be integrated to model phenomena at the application layer time scale. Specifically, we show how a general semi-hidden Markov model can be used to capture the dynamics of the service session over time as well as the users behaviours. We demonstrate the use of the model by different examples. Taken together, our model results able to capture fine and coarse grained behaviour in 3D XR traffic.